Pond Street Ledger

Omnibus, Segregated, or Self-Custody: Who Actually Holds a Stock Token, and What Breaks in Each

Three custody arrangements sit behind tokenized equities, and they fail in different ways. A guide to what each one protects, what it costs, and which question to ask an issuer first.

✓ 1593.efrogs.eth2026-11-218 min

Custody is the least glamorous part of a stock token and the part that decides what you own when something goes wrong. The token on your screen is a claim. Whether that claim survives an issuer failure, a venue failure or a court order depends almost entirely on how the underlying share is held, and on whether the register of who holds what lives in the same place as the shares themselves.

There are three arrangements in common use. They are not equally safe, and the safer ones are slower and more expensive. The tradeoff is consistent: the more legal separation you buy between your share and everyone else's balance sheet, the more operational friction, cost and gatekeeping you accept in exchange.

The three shapes

In an omnibus arrangement, a custodian holds one large position in the underlying stock on behalf of all token holders collectively. The custodian's books show the issuer or a nominee as the owner of, say, 400,000 shares. Who among the token holders owns which slice is recorded elsewhere, usually on the chain itself or in the issuer's own ledger. In a segregated arrangement, each client's position is held in a separately identified account at the custodian, so the custodian's own records show your entitlement, not just the pool's. In a self-custody arrangement, the token sits in a wallet whose private key you control, while a regulated entity somewhere upstream still holds the actual share, because a share certificate cannot live on a chain without a legal bridge.

Note what self-custody does and does not mean here. It means you control the transfer of the token. It does not mean you control the share. Every tokenized equity structure in production has a regulated holder of the real security somewhere behind it, because transfer agents and central securities depositories are the places where share ownership is legally recognised. Self-custody of the token is custody of the claim, not of the asset.

Where each one breaks

Omnibus breaks on reconciliation. If the chain-side register and the custodian's position drift apart, through a failed mint, a bug in the issuance contract, or tokens created against shares that were never purchased, nobody at the custodian notices, because the custodian only ever sees the pool total. The shortfall surfaces at redemption, when more claims arrive than there are shares to satisfy them, and the holders who redeem last discover the problem. Omnibus also entangles you with other holders in an insolvency, since the pool is one position and the question of whose share is whose becomes a matter of proving entitlement from a ledger the court may or may not accept.

Segregated breaks on cost and on the onboarding gate. Individually identified accounts require the custodian to know who you are, which means the token cannot move freely to a wallet the custodian has never seen. In practice this produces transfer restrictions, whitelists and a smaller set of people who can hold the thing, which is precisely the property that makes a token useful onchain and precisely the property segregation removes. You are trading composability for a clean claim.

Self-custody breaks on the bridge and on key loss. The wallet protects you from the venue, not from the issuer: if the entity that promised to deliver a share against your token stops answering, your private key secures a claim against nobody. And the ordinary failure mode of self-custody, a lost key, is unrecoverable in a way that a brokerage account is not. Traditional equity custody has a dull, expensive, slow process for proving you are who you say you are and restoring access. A chain has none, unless the issuer builds a forced-transfer function into the token contract, which immediately reintroduces the central party the self-custody holder thought they had escaped.

The question to ask first

The single most useful thing to establish about any stock token is who the custodian's books say owns the underlying share, and what evidence connects that entry to your balance. Not which chain it is on, not which venue quotes it, not whether it trades at the weekend. If the answer is that the custodian shows a nominee holding a pooled position and the per-holder split lives only in a smart contract, then the smart contract is the register of title in every practical sense, and its correctness is your custody arrangement. That may be acceptable. It should be a known choice rather than a discovery.

The second question is what happens to the record during a chain outage or a sequencer failure. If the onchain balance is the authoritative register and the chain stops producing blocks, there is a window in which nobody can prove current ownership, including the issuer. Structures that keep a mirrored off-chain register solve this and pay for it with the possibility that the two registers disagree, at which point the issuer has to decide which one wins. Most documentation says the off-chain record governs. Most marketing implies the opposite.

Why it matters more as venues multiply

A single issuer with a single custodian is a self-contained problem. The moment the same token trades on multiple venues, or the same underlying share backs tokens issued by two different programmes, the custody arrangement determines whether the two can ever net against each other. Omnibus pools at different custodians cannot settle between themselves without moving real shares through the conventional depository, which is the slow leg the whole exercise was meant to avoid. This is the structural reason tokenized equity markets tend to fragment by issuer rather than consolidate by asset.

What to watch

Three disclosures are worth reading when they appear. First, whether the custodian publishes a position attestation and at what frequency, because a monthly attestation against a continuously minting token is a weak control. Second, whether the token contract has a forced-transfer or freeze function and who can call it, since that defines the real limit of self-custody. Third, whether the issuer states which register governs in a dispute. An issuer that answers all three plainly is not necessarily safer, but it is one you can assess.